GIS STUDIES IN RIAU

obmar · GIS STUDIES IN RIAU

Briefing on Authors and Methodology

Deforestation, Forest Degradation, Biodiversity Loss, and CO2 Emissions
in Riau, Sumatra, Indonesia

WWF Indonesia, Remote Sensing Solutions, Hokkaido University

Authors

WWF Indonesia began work in Riau, Sumatra in 1999 and employs around 30 staff in several offices in the province. WWF has been collecting land use and biodiversity data since 2000 and analyzed them for this report. WWF’s large, experienced ground staff and wide network of collaborators among Government, NGOs and industry ensured that all land use data generated from external reports and through remote sensing were thoroughly verified. WWF maintains the most comprehensive satellite image and GIS database available for the Province of Riau.

RSS and Prof. Dr. Florian Siegert of Munich University have over 15 years experience in the region especially in Indonesia’s Borneo and Sumatra. Current research projects in Indonesia deal with various aspects of tropical deforestation, fire and peat oxidation. Together with Prof. Susan Page und Prof. Jack Rieley, Siegert has been focusing on peat science since 1996 (EU research programmes FP4. FP 5. and FP 6.). They pioneered some of the basic science on peat emissions that everybody is trying to apply today. More information on Siegert’s work is available at www.rssgmbh.de. Selected publications include:
• Rieley, J.O. and Page, S.E. (editors) (2005) Wise Use Guidelines for Tropical Peatlands. Alterra, Wageningen, The Netherlands. 237 p. ISBN 90327-0347-1 http://www.restorpeat.alterra.wur.nl/p_frameset.htm
• Aldhous Peter, (2004). Borneo is burning. Nature, VOL 432, 144-146. (Peter Alhous reported on the work of Prof. Siegert/Prof. Rieley)
• Siegert, F., G. Rücker, A. Hinrichs & A. Hoffmann (2001). Increased fire impacts in logged over forests during El Niño driven fires. Nature, 414, 437-440
• Page S. E., Siegert F., Rieley J. O., Boehm H-D.V.and A. Jaya (2002). Carbon released during peatland fires in Central Kalimantan, Indonesia in 1997. Nature, 420, 61-65

Hokkaido University is the top Japanese University in terms of peat science, studied peat ecosystems for over 100 years in Japan. Prof. Dr. Hatano has over 12 years experience studying Kalimantan’s peat ecosystem as part of a large joint 10 year project between Indonesian-Japanese team of scientists: “Environmental conservation and land use management of wetland ecosystem in Southeast Asia (1997-2006)” funded by JSPS (Japan Society for the Promotion of Science) and co-authored around 100 publications and proceedings. Detailed carbon flux studies with hundreds of soil and canopy emission and absorption CO2 sensors have generated some of the most fundamental baseline data on such eco-systems available to date.

Thanks to Hokkaido University and RSS and their Indonesian colleagues there is no other peat ecosystem in the region that is better studied than Central Kalimantan’s Sebangau National Park and Mega Rice Project.

Study Area

Our deforestation, fire and CO2 emissions study focused on the ca. 8.3 million-hectare mainland of the province of Riau in central Sumatra, Indonesia. More-detailed forest degradation and forest replacement analysis focused on WWF-Indonesia Riau Programme’s 4.5 million-hectare Tesso Nilo-Bukit Tigapuluh–Kampar Conservation Landscape (TNBTK Landscape), covering ca. 55% of Riau’s mainland.

Deforestation and Driver Analysis

We analyzed deforestation in Riau over the last quarter century, between 1982 and 2007. We determined drivers of deforestation, identifying which land covers had replaced the natural forest that had been cleared and which land use zones has seen such changes. Finally, we built two scenarios to predict deforestation between 2007 and 2015, the year until which Riau’s new proposed land use plan is planned to be valid: “Business as usual” and “Full implementation of draft Riau Land Use Plan 2015”.

We defined “forest” as area with original natural forest with a crown cover of more than 10% (following FAO’s definition of forest). We did not include plantations such as acacia and oil palm plantations under the term “forest.” We also did not include forest re-growth in “forest.” We mapped “forest-non forest” cover for the years 1982, 1988, 1996, 2000, 2002, 2004, 2005, 2006, and 2007 for Riau’s 8.3 million-hectare mainland based on information from World Conservation Monitoring Centre, UNEP, Indonesian Ministry of Forestry, and WWF analysis of Landsat images 2000-2007. We distinguished forests on peat versus non peat soils based on a delineation of peatland in Riau by Wetlands International.

We created a detailed land cover GIS database for the “Tesso Nilo - Bukit Tigapuluh - Kampar Landscape”, distinguishing up to 50 land cover classes on dozens of Landsat TM/ETM images and one IRS image for four periods: 1990, 1995, 2000, and 2005. For 2007, we did the same very detailed land cover analysis for the whole mainland of Riau. Images were analyzed on screen with the minimum mapping unit fixed at ca. 50 ha. Land cover was digitized at a scale of 1:90,000. The accuracy of on-screen land cover interpretations was confirmed through frequent field verifications. A comprehensive database with GPS locations and photos of all field verification sites was compiled. The highly labor intensive work of on-screen digitizing and frequent field verifications, complemented solid supplementary data for land use (concession maps, etc.) of the WWF analysis from high resolution images sets it apart from most other studies who use often very inaccurate automated analyses from low resolution images. The GIS expert who conducted this analysis has over 20 years of experience in surveying vegetation in Sumatra.

We distinguished natural forests as dry lowland, peat swamp, swamp and mangrove forest, and divided each forest type into four classes: rather closed canopy (crown cover (cc) > 70%), medium open canopy (70% > cc > 40), very open canopy (40% > cc > 10%), and cleared (cc < 10%). In the absence of a clear definition of what crown cover percentage constitutes forest under the yet-to-be-developed REDD mechanism for Indonesia, we defined “deforestation” as a change of natural forest with “rather dense,” “medium open,” and/or “very open canopy” to any other land cover class (“oil palm plantation”, etc.). Any forest area with a crown cover of <10% was considered deforested.

Forest Degradation Analysis

We analyzed canopy closure change - as one indicator of forest degradation - for the TNBTK Landscape between 1990 and 2007. We defined “forest degradation” as any change in land cover from “rather closed” to “medium open” or “very open canopy”, and from “medium open” to “very open canopy”. Therefore, any negative change in crown cover between 100% and 10% was considered degraded.

Fire Analysis

Fire occurrence was analyzed using two different low resolution satellite sensors, 1.) the National Oceanic and Atmospheric Administration Advanced Very High Resolution Radiometer (NOAA AVHRR) and 2.) the Moderate Resolution Imaging Spectroradiometer (MODIS). Both systems are well established to detect active burning fires, so called “hotspots”, at a spatial resolution of 1 km in tropical regions. To estimate the fire affected area each recorded hotspot coordinate was converted to an area of 1 km² equivalent to the approximate spatial resolution of the sensor. There is a good correlation between burnt areas determined from hotspots and burnt areas derived from high resolution Landsat imagery. The area estimate is conservative, because the burnt area is often underestimated by the hotspot approach: 1.) fires are detected only once or twice a day and rapidly spreading fires escape recording, 2.) smoke from the fire often impedes the detection of hotspots and 3.) ground fires in forests are not hot enough to be detected from space. Areas of overlapping hotspots were considered to be burnt only once.

We compared successive years of land cover change and fire occurrence. All hotspots of a specific period were superimposed on the respective forest cover map.

CO2 Emissions Estimate

CO2 emissions caused by deforestation, forest degradation, peat decomposition and peat fires and carbon sequestration by the growth of acacia and oil palm plantations that replaced natural forests were calculated for the Province of Riau over 17 years from 1990 to 2007. The analysis was based on Landsat image analysis, the only available historical record with high image resolution that allows such analysis. CO2 emissions caused by land use change were approximated following a Stock-Difference Method that estimates the difference in total biomass carbon stock at time t2 and time t1, as described in the 2006 IPCC Guidelines for National Greenhouse Gas Inventories. The values for biomass carbon were based on research conducted within the European Space Agency funded GMES Forest Monitoring program and a thorough literature review on biomass measurements of various tropical land covers for Indonesian and Southeast Asian forest ecosystems. More detailed data is simply not available. The carbon content of the biomass was set to 50% as recommended by the IPCC Guidelines 2006.

This methodology is an improved approach compared to TIER 1 level which uses IPCC standard values as emission factors. Emission factors and carbon change assessment follow the TIER 2 level, which uses country-specific forest biomass value. Remote sensing was used extensively to assess area changes for all land covers and forest degradation as requested for TIER 3.

The biomass of different land covers was determined by calculating the median values of all published biomass values for a specific type. The biomass within specific forest types may vary regionally due to different growth conditions. These regional variations are to some extent contained in the statistical analysis of the published biomass measurements. More precise calculations of the variation in biomass within single land cover types would require extensive field measurements, forest inventory data and permanent sample plots in areas which are very difficult to access on the ground. In addition, the resolution of the Landsat satellite imagery is too low for more detailed assessments. These would require high resolution aerial imagery and 3D LIDAR data. For this region no other satellite data are available for before 2000, assessments based on higher resolution data are therefore impossible.

Emissions were calculated assuming that all carbon lost through deforestation and forest degradation was released into the atmosphere. This assumption may lead to a slight overestimation, because an unknown amount of the harvested biomass was converted into furniture or paper that may remain permanently as such.

CO2 emission estimates followed state of the art methods. CO2 emissions from decomposing peat soil (peat oxidation) were based on a thorough literature review and long-term measurements of the University of Hokkaido in Borneo. CO2 emissions from burning peat were based on published data and own long-term scientific studies (EUTROP, STRAPEAT and RESTOREPEAT projects). Emissions from forest biomass burning were not considered, because there are virtually not data available on fuel loads, burn intensities, degree of fire damage and many other important parameters. We decided to exclude potential emissions of methan from pristine and degraded peatlands, because emissions are low and there are only few measurements available (Rieley, J.O. and Page, S.E. (editors) (2005))

All calculations had to rely on assumptions and simplifications. Several sources of uncertainty lead to a propagation of errors. We did not add error margins to our estimations, as the level of error of our calculation component is not precisely quantifiable. Each component of our calculations contributes to the total uncertainty. However, considering and reflecting on all errors we are convinced that the order of magnitude of the emissions estimate is correct.

Undisturbed peat swamp forests and tropical peatlands store and sequester huge amounts of carbon. The biomass per hectare is 10-15 times higher than that of the forest biomass growing on the peat. Peat soils release huge amounts of CO2 when deforested or burnt. Therefore it is important to include the peatland ecosystem in the emissions analysis. If the vegetation cover is removed from peatlands the carbon balance in peat soils is affected twofold: 1.) carbon sequestration by peat forming plants is stopped and 2.) the peat soil starts to emit CO2 due to the decomposition of soil organic matter (mainly peat decomposition). Sources of CO2 emissions are also autotrophic respiration by roots and above ground parts of the vegetation cover. Peat decomposition can occur naturally when the hydrology of the peat layer is disturbed. The carbon loss from peat decomposition usually exceeds the carbon sequestration of peat. The most dramatic emissions occur when the land use changes lead to burning of the peat. Both, decomposition and fire are induced by human intervention such as drainage and land clearing.

If peatlands are developed for agriculture and plantations they have to be drained to make the peat soil aerobic and the water less acidic. Then the drained the peat layer undergoes quick decomposition through oxidation by microbial activity. Extended drought periods, as they regularly occur during El Niño episodes, lead to very low water tables (often more than 2 meter below the surface) und thus accelerate the oxidation of the dry peat substrate. At the same time the peat becomes very susceptible to burning.

To estimate CO2 emissions by peat decomposition, we determined emission values for different land covers that had replaced the 691,733 ha of peatland forests in the TNBTK Landscape between 1990 and 2007. Peat decomposition is closely related to drainage, i.e. the average depth of the water table below the peat surface. To estimate emissions we correlated measured peat decomposition values to different drainage regimes and land cover classes based on literature reviews and results of long-term studies in Sarawak and Central Kalimantan. The emission values were averaged and assigned to the respective land cover conversion cases. Only published measurements that were taken in a comparable region, ecosystem and covering a time period of at least one year were selected to compensate for seasonal variations in ground water levels. For established plantations CO2 emissions were considered cumulative.

We analyzed the spatio-temporal occurrence of hotspots on peatlands between 1997 and 2007. All hotspots were superimposed on the peatlands map published by Wetlands International and converted into burnt areas. To estimate CO2 emissions caused by fires, we considered different scenarios of peat consumption by fire. For El Niño years, with their intense droughts the water table is usually lower than 1.50 meters. Thus burning is more intense and we assumed that 50 cm of peat burnt away. For normal years, we assumed that 15 cm of peat burnt away on the average. In total a peat volume of 6.314 km³ burnt having a carbon content of 0.379 Gt C. Peat fires in Riau could have released as much as 1.39 Gt CO2 between 1997 and 2007.

We did not estimate potential sequestration of forests. Whether pristine forests sequester carbon, are neutral or even emit CO2 is still under discussion in the scientific community. A recent publication states that primary tropical rain forest sequester CO2, and thus are a carbon sink. However, the prevalent opinion among scientists is that primary rainforests are a climax vegetation, meaning a plant community that is in equilibrium with its environment, i.e. carbon sequestration equals carbon emissions. We also did not estimate CO2 emissions from burning of above-ground biomass, because of too many uncertainties, i.e. amount of fuel, burn severity, forest biomass etc. Therefore, the calculated net emission can be considered as conservative, i.e. an underestimate. Sequestration by coconut and rubber plantations was also excluded.

CO2 emissions from peat (organic matter) decomposition (t CO2/ha/year) related to different land covers. AG : Agriculture, PL : Plantation, CB: Cleared or Burnt forest (no trees standing).
Land Cover Class Land Use Ave.
Drainage Mean Median SD Max. Min.
Acacia plantation AG+PL 53 85 84 41 165 5
Oil palm plantation AG+PL 53 85 84 41 165 5
Small holder oil palm plantation AG+PL 53 85 84 41 165 5
Cleared land CB 21 29 26 9 48 22
"Waste" land CB 21 29 26 9 48 22
Other land covers CB 21 29 26 9 48 22